Hindawi Publishing Corporation ISRN Pharmaceutics Volume 2013, Article ID 818364, 8 pages http://dx.doi.org/10.1155/2013/818364 Research Article In Silico Prediction of Interactions between Site II on Human Serum Albumin and Profen Drugs Hideto Isogai and Noriaki Hirayama Basic Medical Science and Molecular Medicine, Tokai University School of Medicine, 143 Shimokasuya, Isehara, Kanagawa 259-1193, Japan Correspondence should be addressed to Noriaki Hirayama; [email protected] Received 21 January 2013; Accepted 6 February 2013 Academic Editors: R. A. Caceres, A. Ghosal, and H. Sah Copyright © 2013 H. Isogai and N. Hirayama. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Since binding of a drug molecule to human serum albumin (HSA) significantly affects the pharmacokinetics of the drug, it is highly desirable to predict the binding affinity of the drug. Profen drugs are a widely used class of nonsteroidal anti-inflammatory drugs and it has been reported that several members of the profen class specifically bind to one of the main binding sites named site II. The actual binding mode of only ibuprofen has been directly confirmed by X-ray crystallography. Therefore, it is of interest whether other profen drugs are site II binders. Docking simulations using multiple template structures of HSA from three crystal structures of complexes between drugs and HSA have demonstrated that most of the currently available profen drugs should be site II binders. 1. Introduction center. Based on the reliable crystal structures, structure- based approaches are possible. The molecular docking meth- Human serum albumin (HSA) which is the most abun- ods, in particular, which have been largely improved recently dant plasma protein binds vast array of chemically diverse can be applied to predict the interaction modes of drugs exogenous and endogenous molecules [1]. Binding of a drug and the binding sites in atomic detail comparable to the molecule to HSA results in increased solubility in plasma, experimental results. decreased toxicity, and protection against oxidation of the Since nonsteroidal anti-inflammatory drugs (NSAIDs) bound molecule. Since HSA binding is one of the important are among the most commonly used medications [4], the factors which determine the ADME properties of the drugs, drug-drug interactions involving NSAIDs are important it is highly desirable to know the binding affinity of drugs issues in many drug discovery projects. In particular, 2-aryl in order to avoid undesirable drug-drug interactions. There propionic acids (profen drugs) are a widely used class of are two approaches to predict protein-ligand interactions. NSAIDs and represent a significant share of the pharma- Ligand-based approaches mostly use quantitative structure- ceutical market. In Figure 1(a),chemicalstructuresofeleven activity relationships (QSARs) which are based on chemical typical profens are shown. The binding modes of flurbiprofen structures and physicochemical properties of a series of com- [5], ibuprofen [3], ketoprofen [6], naproxen [7], pranoprofen pounds whose HSA binding affinities have been measured [8],andsuprofen[9] have been experimentally suggested so [2]. High-resolution crystal structures of HSA complexed far, albeit the exact binding mode of ibuprofen only was with various molecules have shown that there are two main confirmed by X-ray crystallography. The binding site of these binding sites named sites I and II [3]. As site I is large and six profen drugs is reported to be site II. It is particularly flexible multichamber, a variety of different molecules can interesting to confirm whether all of the eleven profen drugs bind to site I. On the contrary, ligands binding to site II are including ibuprofen bind to site II. Therefore, in this study we usually aromatic carboxylic acids with a negative charged have predicted whether these profen drugs can actually bind groupatoneendofthemoleculeawayfromahydrophobic to site II by docking simulations. 2 ISRN Pharmaceutics The results have shown that docking simulations can versus those in the crystal structure. The rmsd valuesA) ( ˚ satisfactorily predict the binding of site-II-specific drugs to obtained by redocking are 0.28, 0.70, 0.46, 0.31, 0.95, and 0.58 site II. The docking simulations undertaken in a similar way for 1E7A, 2BXA, 1BXE, 2BXF, 2BXG, and 2BXH, respectively. have demonstrated that all of the profen drugs shown in Since prediction within rmsd of 2.0 Aisheldasthepassing˚ Figure 1(a) canbindtositeII. standard, the results indicate that ASEDock is suitable for the HSA-drug system. 2. Methods In this study, three template crystal structures of 1E7A, 2BXA,and1BXEwhichgavethesmallestrmsdsinthe 2.1. Docking Simulations. All docking simulations were redocking simulations were selected as template structures. undertaken by use of software ASEDock10 [ ]. ASEDock is Asthelargestandthesmallestdrugsinsizeareboundto based on unique concept of ASE model and ASE score. Since the HSA molecule in these structures, the consensus analysis ASEDock is free from any bias except for shape, it is a very using these structures as templates is expected to take the robust docking method. During refinement, backbone atoms induced-fit effect into account to some extent and should be of the HSA molecules were fixed and the positions of all other applicable to docking simulations between HSA and a variety atoms were optimized. Docking results were evaluated by an of different molecules. efficiency index defined as follow: 2.3. Selection of Compounds for Evaluation. Since our pur- dock EI = , (1) pose is to predict compounds which likely bind to site II, we dock HA must evaluate whether docking simulations can distinguish where HA denotes the number of nonhydrogen atoms in a site II binders from other compounds. For this evaluation, = + + we prepared three sets of compounds. The first set of 1,444 ligand and dock ele vdw strain.Here, ele and vdw mean electrostatic and van der Waals interaction energies, compounds mainly consists of drug molecules clinically applied in Japan now. The active metabolites of the drugs respectively, between HSA and ligand molecules. strain refers to the difference between the conformation energy ofa which have been reported in the literatures are also included docked ligand and that of the minimum energy confor- in this dataset (DS1). Six drug molecules complexed with HSA in the crystal structures and nine additional compounds mation nearest to the docked-ligand structure. EI dock corresponds to ligand efficiency [11]. The interaction between (Figure 1(c)) whose binding to site II was experimentally biotin and streptavidin is one of the strong noncova- confirmed by NMR [16] were included in the second dataset lent interactions known in nature [12]. The experimentally of DS2. The compounds in DS2, 15 in total, are positive determined interaction energy per nonhydrogen atom is controls whose binding sites have been confirmed by X-ray −1.6 kcal/(mol⋅HA) [13]. Based on this value, the threshold crystallography or NMR. The third dataset (DS3) consists of eleven profen drugs. The binding of only ibuprofen to site II value of EI dock to judge site II binders was arbitrarily set to be −1.5 kcal/(mol⋅HA) in this study. has been crystallographically confirmed. All the compounds A software system MOE (Molecular Operating Environ- in DS2 and DS3 are included in DS1. ment) [14] was used throughout this study. All the calcu- lations were performed on a DELL PC workstation T7500 3. Results and Discussion equipped with two Intel Xeon X5690 processors and 128 GB memory. The docking simulations between three HSA template struc- tures and all molecules in DS1 have been undertaken. 2.2. Selection of Docking Templates. Selection of suitable The docking results were sorted in descending order of template crystal structures is essential for docking simula- EI dock . The docked structure with the lowest EI dock tions. When we started our study, six crystal structures of value for each molecule was selected. The enrichment curve complexes between HSA and different drug molecules were of the site II binders in DS2 is illustrated in Figure 2.The available. The PDB (Protein Data Bank [15]) codes are 1E7A enrichment rate at the −1.5 kcal/(mol⋅HA) cut is 5.67. The (2.20, 0.248), 2BXA (2.35, 0.230), 2BXE (2.95, 0.226), 2BXF rapid enrichment indicates that EI dock is a reasonable (2.95, 0.215), 2BXG (2.70, 0.234), and 2BXH (2.25, 0.227). criterion to judge the site II binders. 13 drugs out of 15 The two numbers in the parentheses denote the resolution site II binders were included in 221 compounds whose − ⋅ and the factor, respectively. Since these six drug molecules EI dock is lesser than 1.5 kcal/(mol HA). The EI dock are relatively variable in chemical structure as shown in values for indomethacin and clotrimazole were −1.09 and Figure 1(b), the crystal structures reveal descent induced-fit −0.64 kcal/(mol⋅HA), respectively, and they were not judged effect around drug-binding sites. In this study, consensus- to be site II binders by the current threshold value of based approach was taken in order to consider the induced-fit EI dock . Indomethacin is regarded as a site I binder in a effect by using multiple templates for docking simulations. recently published paper [17]. Clotrimazole is relatively Redocking simulations for these six structures have been bulkier than other 13 drug molecules and it might well be undertaken for the validation of the algorithm of ASEDock that the molecule cannot be properly accommodated at site II against the HSA-drug system. Docking accuracy is usually in the three template structures used in this study. However, measured by the root-mean-square deviation (rmsd) of non- from the practical point of view, the enrichment rate of 5.67 hydrogen atom positions in the predicted ligand structure seems to be good enough to judge the site II binders.